RVC-Boss/GPT-SoVITS · error · NO_PROMPT_ERROR

prompt_text cannot be empty when using SoVITS_V3

Error message

prompt_text cannot be empty when using SoVITS_V3

What it means

Raised as NO_PROMPT_ERROR when the inference call has no prompt text while the loaded model is SoVITS v2-Pro (configs.use_vocoder is True). The v2Pro vocoder path conditions generation on prompt text tokens; without them the vocoder has no linguistic anchor for the reference, so inference is refused rather than allowed to produce garbage.

Source

Thrown at GPT_SoVITS/TTS_infer_pack/TTS.py:1120

            print(i18n("当开启并行推理模式时，SoVits V3/4模型不支持分桶处理，已自动关闭分桶处理"))
            split_bucket = False
        else:
            print(i18n("分桶处理模式已关闭"))

        # if fragment_interval < 0.01:
        #     fragment_interval = 0.01
        #     print(i18n("分段间隔过小，已自动设置为0.01"))

        no_prompt_text = False
        if prompt_text in [None, ""]:
            no_prompt_text = True

        assert text_lang in self.configs.languages
        if not no_prompt_text:
            assert prompt_lang in self.configs.languages

        if no_prompt_text and self.configs.use_vocoder:
            raise NO_PROMPT_ERROR("prompt_text cannot be empty when using SoVITS_V3")

        if ref_audio_path in [None, ""] and (
            (self.prompt_cache["prompt_semantic"] is None) or (self.prompt_cache["refer_spec"] in [None, []])
        ):
            raise ValueError(
                "ref_audio_path cannot be empty, when the reference audio is not set using set_ref_audio()"
            )

        ###### setting reference audio and prompt text preprocessing ########
        t0 = time.perf_counter()
        if (ref_audio_path is not None) and (
            ref_audio_path != self.prompt_cache["ref_audio_path"]
            or (self.is_v2pro and self.prompt_cache["refer_spec"][0][1] is None)
        ):
            if not os.path.exists(ref_audio_path):
                raise ValueError(f"{ref_audio_path} not exists")
            self.set_ref_audio(ref_audio_path)

View on GitHub (pinned to d523079fc0)

Solutions

1. Supply the transcript of the reference audio as prompt_text (and matching prompt_lang) whenever a v2Pro/vocoder model is loaded.
2. If you must run without prompt text, switch back to a non-Pro sovits model (v1/v2/v3/v4 checkpoint without the Pro vocoder path).
3. In API wrappers, make prompt_text a required field when configs.use_vocoder is True and validate before dispatch.

Example fix

# before
audio = handler.run(text="你好", text_lang="zh", ref_audio_path="ref.wav")  # NO_PROMPT_ERROR

# after
audio = handler.run(
    text="你好", text_lang="zh",
    ref_audio_path="ref.wav",
    prompt_text="这是参考音频的文字内容", prompt_lang="zh",  # required for v2Pro
)

if handler.configs.use_vocoder and not (prompt_text or "").strip():
    raise ValueError("prompt_text is required for SoVITS v2Pro models")

def requires_prompt_text(handler) -> bool:
    """v2Pro/vocoder models cannot run without the reference transcript."""
    return bool(getattr(handler.configs, "use_vocoder", False))

When it happens

Trigger: Calling infer_batch (or the TTS.run wrapper) with prompt_text=None or "" while configs.use_vocoder is True — i.e. a v2Pro sovits model is loaded and the caller tries the 'ref_free' (no prompt text) mode that works for v1/v2.

Common situations: User upgrades to a v2Pro model but keeps an old script/preset that omitted prompt text; webui opened in free-reference mode with a Pro model selected; API request built without prompt_text field.

Related errors

• 参考音频在3~10秒范围外，请更换！
• ref_audio_path cannot be empty, when the reference audio is
• {ref_audio_path} not exists
• 请输入有效文本
• 参考音频在3~10秒范围外，请更换！

AI-assisted analysis of RVC-Boss/GPT-SoVITS@d523079fc0 (2026-08-15). Data as JSON: /api/errors/3a63d4459dfa4dc0. Report an issue: GitHub.